Automated Video Pipelines

// Prompt sent to OpenAI GPT-4 API
{
  "model": "gpt-4",
  "messages": [
    {
      "role": "system",
      "content": "You are a YouTube content strategist for a tech news channel. Generate 1 video topic based on trending AI news. Return JSON with: title, angle, target_audience, estimated_length_seconds, 5 keywords."
    },
    {
      "role": "user",
      "content": "Today's trending topics: OpenAI GPT-5 rumors, AI regulation in EU, Runway Gen-4 launch, Apple Vision Pro AI features"
    }
  ]
}

{
  "title": "Runway Gen-4: The Future of AI Video",
  "total_duration": 90,
  "scenes": [
    {
      "scene_number": 1,
      "narration": "A new era of AI video generation has arrived. Runway just released Gen-4, and it changes everything.",
      "visual_prompt": "Futuristic digital landscape with glowing neural networks transforming into video frames, cinematic lighting, 4K",
      "duration": 8,
      "transition": "fade_in"
    },
    {
      "scene_number": 2,
      "narration": "Gen-4 introduces multi-shot consistency — characters and scenes now maintain their look across an entire video.",
      "visual_prompt": "Split screen showing AI-generated character appearing identical across four different scenes, clean interface design",
      "duration": 10,
      "transition": "cut"
    }
  ]
}

// Generate images for all scenes
const generateSceneImages = async (scenes) => {
  const images = [];
  for (const scene of scenes) {
    const response = await openai.images.generate({
      model: "dall-e-3",
      prompt: scene.visual_prompt + " --style cinematic, consistent color palette, 16:9 aspect ratio",
      n: 1,
      size: "1792x1024",
      quality: "hd"
    });
    images.push({
      scene_number: scene.scene_number,
      url: response.data[0].url,
      revised_prompt: response.data[0].revised_prompt
    });
  }
  return images;
};

// Send each image to Runway Gen-3 for video generation
const generateVideoClip = async (imageUrl, motionPrompt) => {
  const response = await fetch('https://api.runwayml.com/v1/image-to-video', {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${RUNWAY_API_KEY}`,
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      image_url: imageUrl,
      prompt: motionPrompt,
      duration: 5,
      aspect_ratio: "16:9",
      motion_intensity: 0.6
    })
  });
  return response.json();
};

# Concatenate video clips with crossfade transitions
ffmpeg -i clip1.mp4 -i clip2.mp4 -i clip3.mp4 -i clip4.mp4 -i clip5.mp4 \
  -filter_complex "
    [0:v][1:v]xfade=transition=fade:duration=0.5:offset=4.5[v01];
    [v01][2:v]xfade=transition=fade:duration=0.5:offset=9[v012];
    [v012][3:v]xfade=transition=fade:duration=0.5:offset=13.5[v0123];
    [v0123][4:v]xfade=transition=fade:duration=0.5:offset=18[vout]
  " -map "[vout]" output_no_audio.mp4

# Merge video with voiceover and background music
ffmpeg -i output_no_audio.mp4 -i voiceover.mp3 -i bgmusic.mp3 \
  -filter_complex "[1:a]volume=1.0[voice];[2:a]volume=0.15[music];[voice][music]amix=inputs=2[aout]" \
  -map 0:v -map "[aout]" -shortest final_video.mp4

// Upload to YouTube using the Data API v3
const uploadToYouTube = async (videoPath, metadata) => {
  const youtube = google.youtube({ version: 'v3', auth: oauthClient });
  const response = await youtube.videos.insert({
    part: 'snippet,status',
    requestBody: {
      snippet: {
        title: metadata.title,
        description: metadata.description,
        tags: metadata.tags,
        categoryId: '28' // Science & Technology
      },
      status: {
        privacyStatus: 'public',
        publishAt: metadata.scheduledTime,
        selfDeclaredMadeForKids: false
      }
    },
    media: {
      body: fs.createReadStream(videoPath)
    }
  });
  return response.data;
};

const runPipelineStage = async (stageName, stageFn, input, maxRetries = 3) => {
  for (let attempt = 1; attempt <= maxRetries; attempt++) {
    try {
      const result = await stageFn(input);
      await validateOutput(stageName, result);
      console.log(`[${stageName}] Completed on attempt ${attempt}`);
      return result;
    } catch (error) {
      console.error(`[${stageName}] Attempt ${attempt} failed:`, error.message);
      if (attempt === maxRetries) {
        await notifyHuman(stageName, error);
        throw new Error(`Pipeline halted at ${stageName} after ${maxRetries} attempts`);
      }
      await sleep(attempt * 5000); // Exponential backoff
    }
  }
};